docker compose up -dOpen https://localhost, accept the self-signed cert, and sign in:
- Username:
@demo:localhost - Password:
demo
Two chats are waiting: one with @brute:localhost, one with @echo:localhost.
The stack ships with qwen2.5:0.5b pulled at build time so the first docker compose up just works. This is a tiny model (0.5B params) — fine for verifying the wiring but not much use for real conversation.
You probably want something like llama3.2:latest. Pull it into the running Ollama container:
docker compose exec ollama ollama pull llama3.2Then update the model name in agents/brute/config.ru:
llm = Brute::Agent.new(
provider: :ollama,
model: "llama3.2:latest", # <-- change this
...
)Rebuild brute (docker compose up --build brute) and you're set.
bin/generate-agent <name>Or manually:
cp -r ./agents/brute ./agents/<name>- updateDockerfile,config.ru,agent_card.yml- Generate tokens:
openssl rand -hex 32(one for AS_TOKEN, one for HS_TOKEN) docker/synapse/appservices/<name>.yml- copybrute.yml, substitute name + tokensdocker/synapse/homeserver.yaml- add- /data/appservices/<name>.ymltoapp_service_config_filesdocker-compose.yml- add service block, add to bootstrapdepends_on+DEMO_AGENTS
Then edit agents/<name>/config.ru with your logic and:
docker compose up --build